← Clinical Trials
Breast Cancer Phase 3 Randomized NCT00553358

NeoALTTO: Complete Statistical Analysis of Lapatinib and Trastuzumab in Breast Cancer

An independent statistical analysis of the randomized phase 3 NeoALTTO trial evaluating lapatinib, trastuzumab, and their combination, with pathological complete response as the registered primary endpoint and longer-term event-free survival, overall survival, and landmark analyses reported in the registry.

Trial status: Completed  ·  Enrollment: 455  ·  Primary completion: May 27, 2010
Scope of this record

This page separates reported trial results from statistical interpretation. Numerical results are taken from the ClinicalTrials.gov record. This page provides an independent statistical analysis and educational interpretation of publicly reported results. ClinicalTrials.gov provides the official trial registry record. View the ClinicalTrials.gov record.

1. Trial at a Glance

NeoALTTO was a randomized phase 3 oncology trial evaluating three treatment strategies involving lapatinib and trastuzumab, with pathological complete response at the time of surgery as the registered primary endpoint. The trial enrolled 455 participants and posted results for 22 outcome measures, including 14 statistical analyses.

455
Enrollment
Phase 3
3
Treatment Arms
Randomized allocation
21.79
pCR Difference
97.5% CI 9.08–34.23
0.878
EFS HR
95% CI 0.57–1.34
FeatureNeoALTTO
Trial nameNeo ALTTO (Neoadjuvant Lapatinib and/or Trastuzumab Treatment Optimisation) Study
PhasePhase 3
ConditionNeoplasms, Breast
DesignRandomized, unmasked, treatment-purpose trial
Enrollment455
Arms3
Primary endpointNumber of Participants With Pathological Complete Response (pCR) at the Time of Surgery
Primary endpoint time frameWeeks 20 to 22
Primary endpoint typeBinary
Lead sponsorNovartis Pharmaceuticals
Trial statusCompleted
ClinicalTrials.govNCT00553358

2. Clinical Question

The registered primary question concerns whether the percentage of participants achieving a pathological complete response at the time of surgery differed between the three randomized treatment strategies. The registry reports two formal pairwise primary analyses: lapatinib alone versus trastuzumab alone, and the lapatinib-plus-trastuzumab combination versus trastuzumab alone.

Population

Participants enrolled in the NeoALTTO phase 3 study with the condition recorded as neoplasms of the breast.

Intervention strategies

Lapatinib 1500 mg; trastuzumab 2 mg/kg; and lapatinib 1000/750 mg plus trastuzumab 2 mg.

Comparator

The registered pairwise primary comparisons use trastuzumab 2 mg/kg as the comparator arm.

Primary question

Does the percentage of participants with pathological complete response differ between the randomized treatment strategies at the time of surgery?

3. Trial Design

01
Randomize455 participants
02
3 ArmsLapatinib, trastuzumab, combination
03
Neoadjuvant treatmentPrimary endpoint at surgery
04
Assess pCRWeeks 20 to 22
05
Long-term follow-upEFS and OS up to approximately year 10
ARM · Lapatinib

Lapatinib 1500 mg

  • Lapatinib
  • Registered dose label: 1500 mg
  • Included as one of the three randomized treatment strategies
ARM · Trastuzumab

Trastuzumab 2 mg/kg

  • Trastuzumab
  • Registered dose label: 2 mg/kg
  • Used as the comparator in the two primary pairwise analyses
ARM · Combination

Lapatinib 1000/750 mg + trastuzumab 2 mg

  • Lapatinib
  • Trastuzumab
  • Registered dose labels: lapatinib 1000/750 mg and trastuzumab 2 mg
Important registry distinction: the ClinicalTrials.gov record identifies the three interventions and their registered arm labels, but it does not provide a complete treatment-schedule narrative. This page therefore does not infer dosing schedules, duration of therapy, surgery procedures, or other treatment details that are not contained in the ClinicalTrials.gov record.

4. Trial Timing and Registry Structure

Start
January 5, 2008
Primary completion
May 27, 2010
Allocation
Randomized
Masking
None
Primary purpose
Treatment
Results posted
Yes; 22 outcome measures and 14 statistical analyses are posted in the ClinicalTrials.gov record.

5. Endpoints

The registered primary endpoint is binary and is assessed at the time of surgery. The registry also reports secondary time-to-event outcomes and analyses examining associations between pathological complete response and subsequent event-free or overall survival.

EndpointRegistry definition / time frameType
Number of Participants With Pathological Complete Response (pCR) at the Time of Surgery Weeks 20 to 22. Pathological complete response is defined as no invasive cancer in the breast or only non-invasive in situ cancer in the breast specimen. Surgical breast and axillary node resection specimens were evaluated for pathologic tumor response according to National Surgical Adjuvant Breast and Bowel Project (NSABP) guidelines, which do not take into account the histological nodal status. Binary
Event-free Survival (EFS) - Events and Censoring From randomization up to approximately year 10 Time-to-event
Overall Survival (OS) - Deaths and Censoring From randomization up to approximately year 10 Time-to-event
Assess Associations Between Locoregional Pathological Complete Response (pCR) and Event-free Survival (EFS) - Median Clinical Follow-up (EFS Landmark Population) Up to year 10 Time-to-event
Assess Associations Between Locoregional Pathological Complete Response (pCR) and and Overall Survival (OS) - Median Clinical Follow-up (OS Landmark Population) Up to year 10 Time-to-event
Landmark definition: for EFS, the landmark population was the subset of the ITT population who had not had an EFS event within 30 weeks after randomization and were still in clinical follow-up. For OS, the landmark population was the subset of the ITT population who were alive and were followed up for overall survival 30 weeks after randomization.

6. Analysis Populations

The primary pCR analyses and the reported EFS and OS treatment comparisons were conducted in an Intent-to-Treat (ITT) Population. The registry definition describes this as all participants randomized to treatment, except for those who withdrew their consent to use any of their data, as permitted by law in certain countries, prior to receiving the relevant treatment.

PopulationRole in the posted analyses
Intent-to-Treat (ITT)Used for the primary pCR analyses and the reported EFS and OS treatment comparisons.
EFS landmark populationITT participants without an EFS event within 30 weeks after randomization who remained in clinical follow-up.
OS landmark populationITT participants who were alive and followed for OS 30 weeks after randomization.

7. Primary Results: Pathological Complete Response

The registry reports two formal pairwise analyses for the single registered primary endpoint. Both use a binomial method, normalized here as an exact / Clopper-Pearson binomial analysis. The effect measure is the percentage of participants with pCR, expressed in the registry as a difference in percentages.

Lapatinib 1500 mg vs Trastuzumab 2 mg/kg

Difference in percentage of participants with pCR

−4.85

97.5% two-sided CI: −17.6 to 8.16   ·   P = 0.3416

Estimate defined as Arm 1 (Lapatinib 1500 mg) minus Arm 2 (Trastuzumab 2 mg/kg).

Clinical Biostats interpretation

The estimate of −4.85 means that the percentage of participants with pCR was estimated to be 4.85 percentage points lower in the lapatinib 1500 mg arm than in the trastuzumab 2 mg/kg arm under the registry's stated comparison.

The estimate is not a relative risk, odds ratio, or hazard ratio. It does not say that the probability of pCR was 4.85% lower in a relative sense; it is a difference in percentages.

The 97.5% two-sided confidence interval extends from −17.6 to 8.16. Thus, the registry estimate has substantial statistical uncertainty and the interval includes zero, the value corresponding to no difference in percentages.

The P = 0.3416 value addresses evidence against the specified null comparison under the binomial testing framework. It does not measure the size, clinical importance, or probability of the observed effect. The confidence interval is more informative about the precision and range of compatible differences.

The analysis was specified as a superiority comparison. Because the endpoint is binary, this analysis does not require the proportional-hazards assumption used later for EFS and OS analyses.

Trastuzumab 2 mg/kg vs Lapatinib 1000/750 mg + Trastuzumab 2 mg

Difference in percentage of participants with pCR

21.79

97.5% two-sided CI: 9.08 to 34.23   ·   P = 0.0001

Estimate defined as Arm 3 (Lapatinib 1000/750 mg + Trastuzumab 2 mg) minus Arm 2 (Trastuzumab 2 mg/kg).

Clinical Biostats interpretation

The estimate of 21.79 means that the combination arm was estimated to have a pCR percentage 21.79 percentage points higher than the trastuzumab 2 mg/kg arm in the posted primary comparison.

This is an absolute difference in percentages, not a relative percentage increase and not a statement that an individual participant was 21.79 percentage points more likely to benefit.

The 97.5% two-sided confidence interval, 9.08 to 34.23, remains above zero. It quantifies uncertainty around the estimated percentage-point difference rather than describing the range of effects that individual patients could experience.

The P = 0.0001 value provides evidence against the null comparison within the posted binomial testing framework. It should not be interpreted as the probability that the null hypothesis is true, nor does it tell us how clinically important a 21.79 percentage-point difference is.

Because two formal pairwise primary analyses are posted for the same binary endpoint, the interpretation should also remain tied to the trial's stated statistical framework. The ClinicalTrials.gov record identifies the hypothesis type as superiority but does not provide an alpha-spending or multiplicity specification for these two pairwise pCR analyses.

8. Primary Endpoint Statistical Methodology

The pCR endpoint is binary: each participant is classified according to whether pathological complete response was present at the time of surgery. The registry reports a binomial method, normalized as an exact / Clopper-Pearson binomial approach.

Conceptual structure
Observed pCR percentage = 100 × (number with pCR / number evaluated)

For a two-group comparison, the reported estimate is expressed as a difference in the percentages between the specified treatment arms. The exact binomial framework avoids relying on a large-sample normal approximation when constructing the interval or testing the binary endpoint.

The registry reports a 97.5% two-sided confidence interval for each primary comparison. This confidence level is part of the reported analysis and should not be silently replaced by a 95% interval.

9. Secondary Results: Event-Free Survival

Event-free survival was assessed from randomization up to approximately year 10. The registry reports Cox proportional-hazards models with the stratification factors entered as strata variables for pairwise comparisons of individual treatment arms with the trastuzumab-alone arm.

ComparisonHR95% CIP-value
Lapatinib 1000/750 mg + Trastuzumab 2 mg vs Trastuzumab 2 mg/kg 0.878 0.57 to 1.34 0.548
Lapatinib 1500 mg vs Trastuzumab 2 mg/kg 1.005 0.66 to 1.52 0.981
How to read these EFS hazard ratios

An HR of 0.878 corresponds to an estimated hazard approximately 12.2% lower than the comparator under the fitted Cox model, because 1 − 0.878 = 0.122. The 95% CI of 0.57 to 1.34 spans 1, so the estimate is compatible with both a lower and a higher hazard relative to the comparator.

An HR of 1.005 is extremely close to 1. The 95% CI of 0.66 to 1.52 is considerably wider than the point estimate and also spans 1.

Neither HR should be read as a probability that a participant experiences an event, nor as a difference in the proportion of participants who remain event-free at a particular time. The Cox model summarizes relative instantaneous event rates over follow-up, subject to its modeling assumptions.

10. Secondary Results: Overall Survival

Overall survival was evaluated from randomization up to approximately year 10. The same stratified Cox modeling framework was used for pairwise comparisons with the trastuzumab-alone arm.

ComparisonHR95% CIP-value
Lapatinib 1000/750 mg + Trastuzumab 2 mg vs Trastuzumab 2 mg/kg 0.788 0.46 to 1.34 0.379
Lapatinib 1500 mg vs Trastuzumab 2 mg/kg 0.962 0.58 to 1.60 0.880

Combination arm: overall survival

HR 0.788

95% CI: 0.46–1.34   ·   P = 0.379

Relative to the trastuzumab 2 mg/kg arm, the estimated hazard ratio was below 1, but the confidence interval includes 1.

Clinical Biostats interpretation

The OS HR of 0.788 means the fitted model estimated a lower instantaneous hazard of death for the combination arm relative to the trastuzumab arm. Numerically, 1 − 0.788 corresponds to an estimated 21.2% lower hazard.

That statement is conditional on the Cox model and should not be converted into a claim that overall survival was 21.2% higher or that an individual patient's probability of survival increased by 21.2%.

The 95% CI of 0.46 to 1.34 indicates substantial uncertainty around the estimated hazard ratio and includes the null value of 1. The P = 0.379 value is a test statistic summary, not a measure of the magnitude of the observed treatment effect.

11. pCR and Event-Free Survival: Landmark Analyses

The registry separately examined the association between locoregional pathological complete response and subsequent event-free survival using a 30-week landmark. These analyses compare participants with pCR versus no pCR within defined landmark populations rather than comparing the randomized treatment arms directly.

Population / comparisonHR95% CIP-value
All subjects in EFS landmark analysis: pCR vs no pCR 0.481 0.31 to 0.73 0.00079
Lapatinib + trastuzumab arm: pCR vs no pCR 0.350 0.16 to 0.71 0.004
Lapatinib arm: pCR vs no pCR 0.532 0.21 to 1.16 0.134
Trastuzumab arm: pCR vs no pCR 0.601 0.28 to 1.20 0.163

For the overall EFS landmark population, the HR of 0.481 indicates an estimated hazard approximately 51.9% lower for participants with pCR than for those without pCR under the fitted model. In the lapatinib-plus-trastuzumab arm, the corresponding estimate was 0.350, while the lapatinib and trastuzumab arm-specific estimates were 0.532 and 0.601, respectively.

Association is not the randomized treatment effect. The landmark pCR analyses compare participants according to whether pCR occurred. Because pCR is an outcome that occurs after randomization, these analyses do not have the same causal interpretation as the randomized treatment comparisons. The 30-week landmark is used to establish a defined population before subsequent EFS follow-up, but it does not turn pCR into a randomized baseline characteristic.

12. pCR and Overall Survival: Landmark Analyses

The registry also reports analogous landmark analyses for overall survival among participants who were alive and followed for OS 30 weeks after randomization.

Population / comparisonHR95% CIP-value
All subjects in OS landmark analysis: pCR vs no pCR 0.366 0.20 to 0.63 0.00041
Lapatinib + trastuzumab arm: pCR vs no pCR 0.223 0.07 to 0.58 0.002
Lapatinib arm: pCR vs no pCR 0.433 0.12 to 1.17 0.125
Trastuzumab arm: pCR vs no pCR 0.414 0.15 to 1.00 0.058

In the overall OS landmark analysis, the estimated HR was 0.366. This corresponds to an estimated 63.4% lower instantaneous hazard of death for the pCR group relative to the no-pCR group under the fitted model. The lapatinib-plus-trastuzumab arm had an HR of 0.223, while the lapatinib and trastuzumab arms had HRs of 0.433 and 0.414, respectively.

Interpretation of the landmark analyses

The association between pCR and later EFS or OS is statistically interesting because the HRs are generally below 1, but it should not be interpreted as evidence that producing pCR through a particular treatment necessarily causes the subsequent survival difference observed in these comparisons.

Participants who achieve pCR may differ from participants who do not in ways that are related to prognosis. The landmark design addresses the timing of classification by conditioning on survival and follow-up at 30 weeks, but it does not recreate the balance of a randomized comparison between pCR and no pCR.

13. Statistical Methods Explained

Why was an exact binomial method used for pCR?

pCR is a binary outcome: participants either meet the registry's pathological definition or they do not. Exact binomial procedures provide confidence intervals and tests for binomial proportions without relying on the normal approximation that can be less reliable with small counts or proportions near the boundaries.

What does a difference of 21.79 mean?

The posted estimate of 21.79 is a difference in percentages, not a ratio. Because the registry defines it as Arm 3 minus Arm 2, a positive value means the estimated pCR percentage in the lapatinib-plus-trastuzumab arm was higher by 21.79 percentage points than in the trastuzumab arm.

Why does the pCR analysis use a different method from EFS and OS?

pCR is measured as a binary endpoint at a defined surgical time point, so a binomial framework is natural. EFS and OS incorporate the timing of events and censoring, so a time-to-event model such as the Cox proportional-hazards model is appropriate for the posted analyses.

What does a hazard ratio of 0.878 mean?

An HR of 0.878 indicates a lower estimated instantaneous event hazard in the first-listed combination arm relative to trastuzumab alone, with the model estimating approximately 12.2% lower hazard. It is not equivalent to a 12.2% improvement in survival probability at a fixed time.

Why is the confidence interval important?

The confidence interval shows the precision of the effect estimate under the specified statistical framework. For example, the EFS HR of 0.878 has a 95% CI from 0.57 to 1.34, which is much wider than the point estimate alone would suggest and includes the null value of 1.

Why should pCR landmark analyses not be treated like randomized comparisons?

Randomization balances treatment assignment at baseline, but pCR is measured after treatment begins. Comparing pCR versus no pCR therefore compares groups defined by a post-randomization outcome. The landmark framework establishes a common 30-week starting point for later follow-up, but it does not restore the causal interpretation of the original randomization.

What does stratification mean in the Cox analyses?

The registry states that the Cox models used the stratification factors as strata variables. Stratification allows the baseline hazard to vary across the strata while estimating a common treatment-associated hazard ratio within the model. The ClinicalTrials.gov record does not specify the names of the stratification factors, so none are inferred here.

14. Understanding the Cox Proportional-Hazards Model

The secondary EFS and OS treatment comparisons and the pCR landmark analyses use Cox proportional-hazards models. Conceptually, the model relates the instantaneous event hazard to the comparison of interest while allowing the baseline hazard to vary over time.

Conceptual Cox model
h(t | X) = h0(t) exp(βX)

The hazard ratio associated with a one-unit treatment indicator is exp(β). A value below 1 indicates a lower modeled hazard for the first comparison group; a value above 1 indicates a higher modeled hazard.

The principal modeling caution is the proportional-hazards assumption: the hazard ratio is most naturally interpreted as a relative hazard that is stable over the relevant time scale. The ClinicalTrials.gov record does not report a formal test or diagnostic for that assumption, so the page does not claim that proportional hazards were demonstrated.

15. Confidence Intervals and P-values

The NeoALTTO results illustrate why an effect estimate, confidence interval, and P-value should be read together rather than separately.

AnalysisEstimateConfidence intervalP-value
pCR: Lapatinib vs trastuzumab−4.8597.5% CI −17.6 to 8.160.3416
pCR: Combination vs trastuzumab21.7997.5% CI 9.08 to 34.230.0001
EFS: Combination vs trastuzumabHR 0.87895% CI 0.57 to 1.340.548
EFS: Lapatinib vs trastuzumabHR 1.00595% CI 0.66 to 1.520.981
OS: Combination vs trastuzumabHR 0.78895% CI 0.46 to 1.340.379
OS: Lapatinib vs trastuzumabHR 0.96295% CI 0.58 to 1.600.880

A P-value is calculated relative to a specified null hypothesis and statistical model. It is not the probability that the null hypothesis is true and it does not measure the magnitude of an effect. Confidence intervals provide a complementary view of uncertainty and are particularly important for the EFS and OS hazard ratios, where the intervals are substantially wider than the point estimates.

16. Stratified Survival Analysis

The registry states that the EFS and OS Cox models included the stratification factors as strata variables. This is an important distinction from simply including every factor as an ordinary covariate.

Why stratify?

A stratified Cox model permits different baseline hazards across the specified strata while retaining the treatment comparison as the principal effect of interest.

What is not reported?

The ClinicalTrials.gov record does not identify the individual stratification factors or their levels, so this page does not reconstruct them from external sources.

Stratification is especially relevant when randomization or trial design has deliberately accounted for prognostic factors. It can help align the analysis with the structure of the randomized comparison without assuming that every stratification factor has the same baseline hazard over time.

17. Safety Results

The ClinicalTrials.gov record reports serious adverse events by treatment arm as affected participants divided by participants at risk. These denominators are reported exactly as reported in the registry.

Treatment armSerious adverse eventsAffected / at risk
Lapatinib 1000/750 mg + Trastuzumab 2 mg 61 participants 61 / 149
Lapatinib 1500 mg 58 participants 58 / 151
Trastuzumab 2 mg/kg 36 participants 36 / 148

These figures describe serious adverse events using the affected/at-risk counts reported by the registry. They should not be substituted for other safety categories, such as overall adverse events or grade-specific adverse events, because those measures are not contained in the ClinicalTrials.gov record.

Safety denominator caution: the registry-reported serious-adverse-event denominators are 149, 151, and 148. The trial-level enrollment is 455. This page reports the registry's arm-specific safety denominators as given and does not infer a different randomized-arm accounting from the enrollment total.

18. What the Primary pCR Results Do — and Do Not — Mean

Statistical interpretation

The combination-versus-trastuzumab primary analysis produced an estimated pCR percentage difference of 21.79, with a 97.5% two-sided CI of 9.08 to 34.23 and P = 0.0001. The estimate is a percentage-point difference, not a hazard ratio or relative risk.

The lapatinib-versus-trastuzumab analysis produced an estimated difference of −4.85, with a 97.5% two-sided CI of −17.6 to 8.16 and P = 0.3416. Its interval includes zero, so the posted estimate is statistically compatible with both a lower and a higher pCR percentage for lapatinib relative to trastuzumab.

Why the confidence intervals matter

The pCR confidence intervals quantify uncertainty around the estimated differences. The combination comparison has an interval entirely above zero, whereas the lapatinib comparison has an interval spanning zero. These intervals describe uncertainty under the specified analysis; they are not prediction intervals for individual patients.

Why pCR and survival should be separated

pCR is a binary endpoint measured at surgery. EFS and OS are time-to-event outcomes extending from randomization. A statistical association between pCR and later survival is therefore conceptually different from the randomized treatment comparison of pCR itself.

19. Multiplicity and Multiple Comparisons

The ClinicalTrials.gov record identifies two formal primary- endpoint analyses for the same pCR endpoint and labels both as superiority analyses. It does not provide an alpha-spending scheme, a formal multiplicity-adjustment method, or a hierarchy governing these two pairwise comparisons.

AnalysisRole in the ClinicalTrials.gov recordReported framework
pCR: Lapatinib vs trastuzumabPrimaryExact / Clopper-Pearson binomial; superiority
pCR: Combination vs trastuzumabPrimaryExact / Clopper-Pearson binomial; superiority
EFS treatment comparisonsSecondaryStratified Cox proportional-hazards model; superiority
OS treatment comparisonsSecondaryStratified Cox proportional-hazards model; superiority
pCR landmark analysesSecondaryStratified Cox proportional-hazards model; association of pCR with survival

Because the ClinicalTrials.gov record does not specify how multiplicity across the primary pairwise comparisons was handled, the page does not impose a correction or reinterpret the reported P-values using an unstated procedure. This is an important limitation when translating multiple formal comparisons into a single overall error-control statement.

20. Missing Data, Censoring, and Analysis Assumptions

The registry provides a specific ITT definition for the posted analyses and identifies censoring as part of the EFS and OS outcome measures. However, the ClinicalTrials.gov record does not describe a detailed missing-data or imputation strategy for the pCR endpoint, nor does it specify the censoring rules used for every possible EFS or OS scenario.

Binary pCR endpoint

The analysis classifies participants according to the registry's pathological definition. The ClinicalTrials.gov record does not provide an additional imputation algorithm for missing pCR assessments.

Time-to-event endpoints

EFS and OS incorporate events and censoring. Cox modeling uses the observed event and censoring structure available for the analysis population.

Proportional hazards

The Cox model requires an interpretable proportional-hazards structure. The ClinicalTrials.gov record does not report a formal diagnostic of this assumption.

Landmark selection

The pCR association analyses use a 30-week landmark, restricting later follow-up to prespecified landmark populations.

21. Why This Trial Matters Statistically

NeoALTTO is a useful teaching case because the same randomized trial contains two very different statistical questions: a binary pathological response endpoint assessed at surgery and time-to-event outcomes followed for approximately a decade. It also illustrates the important distinction between a randomized treatment comparison and an observational-style association defined by a post-randomization response.

ConceptHow it appears in NeoALTTO
RandomizationThree randomized treatment strategies in a phase 3 trial
ITT analysisPrimary pCR and reported EFS/OS treatment analyses use the ITT population
Binary endpointpCR at the time of surgery, assessed during weeks 20 to 22
Exact binomial analysisPrimary pCR comparisons use an exact / Clopper-Pearson binomial method
Confidence interval97.5% two-sided intervals for the primary pCR comparisons and 95% intervals for the Cox analyses
Hazard ratioEFS and OS treatment comparisons and pCR landmark associations
Cox proportional-hazards modelUsed for EFS, OS, and landmark analyses
Stratified analysisStratification factors entered as strata variables in Cox models
Time-to-event endpointsEFS and OS assessed from randomization up to approximately year 10
Landmark analysis30-week landmark used for pCR associations with EFS and OS
Post-randomization variablepCR is analyzed as an exposure-like grouping in later survival association analyses

22. A Worked Reading of the NeoALTTO Evidence

A useful way to read this trial statistically is to move from the primary endpoint outward rather than treating every reported P-value as equivalent.

Step 1 · Primary endpoint

Start with pCR

The registered primary endpoint is a binary measure at surgery, so the exact binomial framework is the natural starting point. The key treatment comparisons are expressed as differences in pCR percentages.

Step 2 · Precision

Read the confidence interval

The 21.79 percentage-point combination estimate has a 97.5% CI of 9.08 to 34.23, while the −4.85 lapatinib estimate has a 97.5% CI of −17.6 to 8.16.

Step 3 · Longer-term outcomes

Move to EFS and OS

EFS and OS are different endpoints requiring time-to-event methods. The posted analyses use stratified Cox models and report hazard ratios with 95% confidence intervals.

Step 4 · Association

Separate pCR association from treatment effect

The landmark analyses ask whether pCR is associated with subsequent EFS or OS. They do not directly estimate the randomized treatment effect because pCR occurs after treatment assignment.

23. Limitations

24. Statistical Methods Explained: Practical Takeaways

Binary versus time-to-event

pCR is a binary outcome assessed at a defined time, whereas EFS and OS preserve information about when an event occurs and whether follow-up is censored.

Difference versus hazard ratio

A pCR difference is expressed in percentage points. An EFS or OS HR compares modeled instantaneous event hazards. They are not interchangeable effect measures.

Primary versus secondary

The registered pCR endpoint is primary. The posted EFS, OS, and landmark analyses are secondary and answer different questions.

Randomized versus landmark comparison

Treatment assignment is randomized. pCR status is not. That distinction is central to interpreting the later pCR-survival analyses.

25. Related Tutorials

Learn more about the methods used in this trial:

26. Related Statistical Calculators

27. Sources

Continue through the Clinical Biostats statistical library

Use the trial's endpoints and methods as a practical route into deeper statistical tutorials, calculators, and clinical-trial methodology.

28. Record Summary

NeoALTTO provides a compact illustration of how different statistical questions arise within one randomized clinical trial. The registered primary endpoint is a binary pathological response measure analyzed using an exact binomial framework, with two formal pairwise comparisons. Longer-term EFS and OS outcomes are analyzed with stratified Cox proportional-hazards models, producing hazard ratios and confidence intervals. Separate 30-week landmark analyses examine associations between pCR and subsequent survival, but those analyses should not be given the same causal interpretation as the randomized treatment comparisons because pCR is a post-randomization outcome.

The most useful statistical reading therefore keeps four elements distinct: the randomized treatment comparison, the endpoint-specific effect measure, the uncertainty surrounding each estimate, and the population from which the estimate was obtained. In NeoALTTO, those distinctions are especially important because pCR, EFS, OS, and pCR-landmark analyses answer related but non-identical questions.

Clinical Biostats methodology: A trial-results page should not merely repeat registry fields. The goal is to reconstruct the statistical story of the trial while clearly separating reported numerical evidence from educational interpretation and avoiding conclusions that the underlying analysis does not support.